In [10]:
import pandas as pd
import numpy as np
# Setup
import os
import pandas as pd
import numpy as np
import h5py
import scipy.io
import string

import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import matplotlib.ticker as ticker
import matplotlib.animation as animation
import matplotlib as mpl
import seaborn as sns


from scapy.all import RawPcapReader
from datetime import datetime
from pathlib import Path
from tqdm import tqdm
from sklearn.preprocessing import MinMaxScaler
from sklearn.manifold import TSNE
from sklearn.preprocessing import LabelEncoder


csv_summary = pd.read_csv('../EDA/SAIDAS/04-csv_summary.csv')

activities = {
  'A': 'Push forward',
  'C': 'Hands up and down',
  'P': 'Reading',
  'S': 'Writing'
}

csv_summary = csv_summary.loc[
  (csv_summary['test'] == 'fine_grained')
  & (csv_summary['slot'] == 'Test')
  & (csv_summary['label'].isin(activities.keys()))]

csv_summary['activity'] = csv_summary['label'].map(activities)
In [11]:
def get_subset(csv_summary,
               monitors=['m1', 'm2', 'm3'],
               environments=['Classroom', 'Office'],
               labels=['A', 'C', 'P', 'S'], 
               samplesize=None,
               random_state=42):
  subset = csv_summary.loc[
    (csv_summary['test'] == 'fine_grained')
    & (csv_summary['slot'] == 'Test')
    & (csv_summary['monitor'].isin(monitors))
    & (csv_summary['environment'].isin(environments))
    & (csv_summary['label'].isin(labels))
  ]
  if samplesize is not None:
    subset = subset.sample(n=samplesize, random_state=random_state)
  subset = subset.reset_index(drop=True)
  return subset

get_subset(csv_summary, monitors=['m1'], environments=['Classroom'], samplesize=10)
Out[11]:
test environment monitor slot csv_file_path filename label nr_packets nr_subcarriers set activity
0 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... C_batch/batch_535.mat C 50.0 242.0 test Hands up and down
1 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... S_batch/batch_606.mat S 50.0 242.0 test Writing
2 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... C_batch/batch_770.mat C 50.0 242.0 test Hands up and down
3 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... C_batch/batch_1054.mat C 50.0 242.0 test Hands up and down
4 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... A_batch/batch_545.mat A 50.0 242.0 test Push forward
5 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... S_batch/batch_109.mat S 50.0 242.0 test Writing
6 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... A_batch/batch_396.mat A 50.0 242.0 test Push forward
7 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... P_batch/batch_646.mat P 50.0 242.0 test Reading
8 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... A_batch/batch_134.mat A 50.0 242.0 test Push forward
9 fine_grained Classroom m1 Test ../Data/fine_grained/Classroom/80MHz/3mo/m1/Sl... S_batch/batch_326.mat S 50.0 242.0 test Writing
In [51]:
def plot_tsne(subset, title='', output_file=f'SAIDAS/08-tsne-output.png', perplexity=30):
  X = np.zeros((len(subset), 50 * 242 * 2))

  for i in range(len(subset)):
    csv_path = subset.iloc[i]['csv_file_path']
    sample_file_path = subset.iloc[i]['filename']
    base_folder = os.path.dirname(csv_path)
    sample_file_path = os.path.join(base_folder, sample_file_path)
    data = scipy.io.loadmat(sample_file_path)['csi_mon']
    real_part = np.real(data)
    imag_part = np.imag(data)
    X[i] = np.concatenate((real_part.flatten(), imag_part.flatten()))

  X = X.astype(np.float32)
  labels = subset['activity'].tolist()

  label_encoder = LabelEncoder()
  numeric_labels = label_encoder.fit_transform(labels)
  class_names = label_encoder.classes_

  colormap = 'tab10'
  cmap = plt.get_cmap(colormap, len(class_names))

  tsne = TSNE(n_components=2, perplexity=perplexity, random_state=42)
  X_embedded = tsne.fit_transform(X)
  x, y = X_embedded[:, 0], X_embedded[:, 1]

  fig, ax = plt.subplots(1, 2, figsize=(14, 6))
  ax[0].scatter(x, y, c=numeric_labels, cmap=cmap, s=5)
  ax[0].set_title(title)
  ax[0].set_xlabel('t-SNE Component 1')
  ax[0].set_ylabel('t-SNE Component 2')
  ax[0].grid(True)
 
  # Polar
  r = np.sqrt(x**2 + y**2)
  theta = np.arctan2(y, x)
  ax[1].scatter(theta, r, c=numeric_labels, cmap=colormap, s=5)
  ax[1].set_title(f'{title} (polar)')
  ax[1].set_xlabel('radius')
  ax[1].set_ylabel('theta')

  plt.tight_layout()

  legend_handles = []
  for i, class_name in enumerate(class_names):
      # Use the same cmap to get the color for each class
      color = cmap(i)
      legend_handles.append(plt.Line2D([0], [0], marker='o', color='w', label=class_name,
                                      markersize=10, markerfacecolor=color))

  plt.legend(handles=legend_handles, loc='upper right', bbox_to_anchor=(1.15, 1), fontsize=8)



  plt.savefig(output_file, dpi=300, bbox_inches='tight')
  plt.show()
In [54]:
subset = get_subset(csv_summary, monitors=['m1'], environments=['Classroom'])
plot_tsne(subset, title='Classroom - Monitor 1', output_file='SAIDAS/08-tsne-classroom-m1.png', perplexity=50)
No description has been provided for this image
In [55]:
subset = get_subset(csv_summary, environments=['Classroom'])
plot_tsne(subset, title='Classroom - Monitors 1, 2 and 3', output_file='SAIDAS/08-tsne-classroom-m1m2m3.png', perplexity=100)
No description has been provided for this image
In [56]:
subset = get_subset(csv_summary)
plot_tsne(subset, title='Classroom and Office - Monitors 1, 2 and 3', output_file='SAIDAS/08-tsne-classroom_office-m1m2m3.png', perplexity=100)
No description has been provided for this image
In [57]:
subset = get_subset(csv_summary, samplesize=5000)
plot_tsne(subset, title='Classroom and Office - Monitors 1, 2 and 3', output_file='SAIDAS/08-tsne-classroom_office-m1m2m3-5000.png', perplexity=100)
No description has been provided for this image